Chapter 12

Norm Topology

[8/100%]
Show
LaTeX
Problem 99

If a Hilbert space H\mathbf{H} is separable, does it follow that the metric space B(H)\mathbf{B}(\mathbf{H}) of operators on it is separable?

?
Remark.
?

If the distance between two operators AA and BB is defined to be ∥A−B∥\left\| A-B\right\|, the set B(H)\mathbf{B}(\mathbf{H}) of all operators on a Hilbert space H\mathbf{H} becomes a metric space.

Problem 100

The set of invertible operators is open. Is the mapping A↦A−1A \mapsto A^{-1} of that set onto itself continuous?

?
Problem 101

Is there an open conjugate class?

?
Remark.
?

The conjugate class of an operator AA is the set of all operators similar to AA (i.e., the equivalence class of AA under the relation of similarity). Conjugate classes do not have to be closed sets, and they do not have to be open (consider A=0A = 0).

Problem 102

If k=1,2,3,…k = 1, 2, 3, \ldots and if k=∞k = \infty, let AkA_k be the two-sided weighted shift such that AkenA_k e_n is en+1e_{n+1} or (1/k)en+1(1/k)e_{n+1} according as n≠0n \neq 0 or n=0n = 0. (Put 1/∞=01/\infty = 0.) What are the spectra of the operators AkA_k (k=1,2,3,…,∞)(k = 1, 2, 3, \ldots , \infty )?

?
Problem 103

Spectrum is upper semicontinuous.

?
Remark.
?

Upper semicontinuity for a set-valued function such as spec⁡\operatorname {spec} has at least two definitions, a metric one and a sequential one. Metric definition: to each open set Λ0\Lambda_0 that includes spec⁡A\operatorname {spec} A there corresponds a positive number ε\varepsilon such that if ∥A−B∥<ε\left\| A-B\right\| < \varepsilon, then spec⁡B⊂Λ0\operatorname {spec} B \subset \Lambda_0. The sequential definition has to be preceded by an auxiliary concept: if {Λn}\left\{ \Lambda_n\right\} is a sequence of sets of complex numbers, then limsup⁡nΛn\operatorname {limsup}_n \Lambda_n is the set of all limit points of those sequences {λn}\left\{ \lambda_n\right\} for which λn∈Λn\lambda_n \in \Lambda_n for each nn. Sequential definition of upper semicontinuity: whenever An→AA_n \to A, it follows that

limsup⁡nspec⁡An⊂spec⁡A. \operatorname *{limsup}_n \operatorname {spec} A_n \subset \operatorname {spec} A.

The two definitions are equivalent.

Problem 104

Is it true that to each operator AA and to each positive number δ\delta there corresponds a positive number ε\varepsilon such that if ∥A−B∥<ε\left\| A-B\right\| < \varepsilon, then ∣r(A)−r(B)∣<δ\left|r(A) - r(B)\right| < \delta? Equivalently: if An→AA_n \to A, does it follow that r(An)→r(A)r(A_n) \to r(A)?

?
Remark.
?

Note that the example of Problem 102 gives no information; in that case the spectral radius is equal to 11 for each term of the sequence and also for the limit.

Problem 105

Is the restriction of spec⁡\operatorname {spec} to the set of normal operators continuous?

?
Remark.
?

If {Λn}\left\{ \Lambda_n\right\} is a sequence of sets of complex numbers, define liminf⁡nΛn\operatorname {liminf}_n \Lambda_n to be the set of all limits of those convergent sequences {λn}\left\{ \lambda_n\right\} for which λn∈Λn\lambda_n \in \Lambda_n for each nn. (Recall that limsup⁡nΛn\operatorname {limsup}_n \Lambda_n was defined as the set of all limit points of not necessarily convergent sequences of this kind; see Problem 103.) The sequence {Λn}\left\{ \Lambda_n\right\} is called convergent (Λn→Λ)(\Lambda_n \to \Lambda ) if and only if

liminf⁡nΛn=limsup⁡nΛn; \operatorname *{liminf}_n \Lambda _n = \operatorname *{limsup}_n \Lambda _n;

in that case the common value deserves to be called lim⁡nΛn\lim_n \Lambda_n. To say that spec⁡\operatorname {spec} is continuous at a particular operator AA means that An→AA_n \to A implies

spec⁡An→spec⁡A. \operatorname {spec} A_n \to \operatorname {spec} A.

Since upper semicontinuity is true for all AA (Problem 103), a necessary and sufficient condition for continuity at AA is that

spec⁡A⊂liminf⁡nspec⁡An. \operatorname {spec} A \subset \operatorname *{liminf}_n \operatorname {spec} A_n.

Recall also that, for normal operators, spectral radius is equal to norm.

Problem 106

If AA and BB are operators such that spec⁡(A+nB)=spec⁡A\operatorname {spec}(A + nB) = \operatorname {spec} A for n=0,1,2,…n = 0, 1, 2, \ldots, does it follow that BB is quasinilpotent?

?
Remark.
?

If small is interpreted in the sense of norm, a perturbation that produces only a small change in the spectrum does not have to be small: if, for instance, A=(α00β)A = \begin{pmatrix} \alpha & 0 \\ 0 & \beta \end{pmatrix} and B=(0100)B = \begin{pmatrix} 0 & 1 \\ 0 & 0 \end{pmatrix}, then spec⁡(A+B)=spec⁡A={α,β}\operatorname {spec}(A+B) = \operatorname {spec} A = \left\{ \alpha , \beta \right\}, but BB is not small; in fact spec⁡(A+nB)=spec⁡A\operatorname {spec}(A+nB) = \operatorname {spec} A for all nn. The answer is clearly yes in some degenerate cases: if, for instance, AA is a scalar, and if spec⁡(A+nB)=spec⁡A\operatorname {spec}(A+nB) = \operatorname {spec} A for just one non-zero value of nn, then spec⁡B=0\operatorname {spec} B = 0.